The final image of the enhanced “Wizard of Oz” film at the Las Vegas Sphere, courtesy of Suvaditya Mukherjee.
Suvaditya Mukherjee stared at Judy Garland’s angelic face, projected on a 160,000 square foot wraparound screen — the size of four football fields.
It was the premiere of the “The Wizard of Oz” at the Sphere, the groundbreaking reimagining of the 1939 classic that opened Aug. 28 at the colossal Las Vegas venue.
He watched as she clicked her ruby slippers and recited: “There’s no place like home.”
Mukherjee didn’t have to be from Kansas to understand.
He remembers, with great clarity, August 3, 2024.
On that date, one year earlier, he stepped off a plane in Los Angeles, feeling extremely disconnected. “It was the first time really being so far away from my parents.”
To be precise: 8,700 miles of disconnection from his native Mumbai, India.
Now, Mukherjee, a master’s student in computer science at the USC Viterbi School of Engineering, is an AI engineer on the enhanced “The Wizard of Oz,” a film that has been called “Hollywood’s biggest AI debut.” It’s estimated to take in more than $1 billion over the next several years — not bad for a movie so old it remembers when movie tickets were less than a quarter.
For Mukherjee, it was “the experience of a lifetime.”
The Wizard of AI
Even though Mukherjee had spent months staring at shots from the movie, he’d never seen the film in its entirety until one night before the premiere.
The project brought together thousands of researchers, programmers, visual effects artists and producers from Magnopus, Google DeepMind, Google Cloud, Sphere Studios and Warner Bros. Discovery to accomplish something that would have been nearly impossible just a few years ago.
Indeed, the original movie, shot in a box-like, 4:3 aspect ratio, might’ve looked like a postage stamp on such a futuristic screen. It required Google DeepMind’s advanced AI models to transform grainy Technicolor frames into stunning 16K resolution imagery.
“When I started thinking about being a machine learning engineer,” Mukherjee said, “I didn’t really think I would apply it to something as real life as entertainment. Particularly something with so much spectacle!”
Suvaditya Mukherjee, a USC Viterbi master’s student, at Google Cloud Next ’25 in Las Vegas.
Working with Oscar winners and veterans who had created effects for films like “Superman” and “The Matrix,” Mukherjee joined Magnopus as part of an elite “strike team” assembled for the final two-month push before the deadline. His role focused on solving some of the trickiest technical challenges — shots that weren’t achieving the highest quality results through the standard pipeline.
The technical hurdles were immense. Using versions of Veo, Imagen and Gemini, the teams developed an AI-based “super resolution” tool to turn tiny celluloid frames from 1939 into ultra-high-definition 16K imagery, then performed “outpainting” — an AI-powered image editing technique that expands an image’s borders — to expand scenes to fill Sphere’s massive canvas.
Using AI to enhance 90% of the film, the team preserved the original performances while adding sharper facial details — freckles on Dorothy’s face, textures on Scarecrow’s burlap skin — and digitally reintegrating characters in scenes where they were initially off camera.
Beyond visual enhancements, the immersive experience includes synchronized 4D effects: 750-horsepower fans to simulate the tornado, haptic seats, fog, scent and even helium-filled drone monkeys flying overhead during key moments. According to Jen Koester, president and COO of Sphere, AI was used not to rewrite the film but “to pull shots together … to create a more seamless and engaging narrative.”
Mukherjee’s wizardry is in optimization and character likeness. Not only did he help artists and engineers compress their generation times, he tackled one of the project’s most delicate challenges: ensuring that AI-generated characters remained faithful to their 1939 counterparts across every frame.
“While our mind can fill in certain gaps (in visual information), the AIs can’t,” Mukherjee explained, describing how rapid movements in the original film — like Ray Bolger’s energetic dances as the Scarecrow — created particular difficulties. “We had to choose the correct frames and then sort of fill in between.”
Midjourney illustration: Mukherjee’s personal yellow brick road led him not to an Emerald City, but a City of Angels — 8700 miles from his native Mumbai, India.
Lost and Found in Oz
Mukherjee’s first week in Los Angeles wasn’t quite eventful as battling flying monkeys or visiting Emerald Cities.
Instead, he and his roommate visited government offices, battled the gas company and struggled to find quality Indian restaurants. In the end, they learned to make their own aloo posto, an Indian delicacy with potatoes and poppy seeds.
But Mukherjee had a plan. And despite being a graduate student specializing in AI, he did it the old-fashioned way.
Over the course of two days, Mukherjee emailed 248 USC professors.
He combed through faculty lists, making an Excel spreadsheet of USC faculty with “any kind of interest in AI and machine learning.” He emailed them all individually.
Of the 10-11 professors who responded, two offered him jobs: Benjamin Nye, director of learning sciences at the USC Institute for Creative Technologies (ICT); and Mark Bolas, a professor at the USC School of Cinematic Arts and a pioneer in mixed reality.
Said Bolas: “I don’t remember the email. What I remember is that I gave him a call to meet, and he was at my office in 20 minutes.”
“I think it just got me really busy,” Mukherjee said of the new jobs, “and that sort of removed the homesickness quite a bit.”
“The serendipity that comes out of being at a campus like USC is really interesting,” Mukherjee added. “One thing led to another.”
In this case, Mukherjee’s path to the Sphere started with a casual conversation with Bolas, for whom Mukherjee works part-time developing a Python course for game developers. He’d seen the project announced at Google Cloud Next in Las Vegas and mentioned his enthusiasm.
Bolas happened to know someone at Magnopus, one of the Sphere’s major partners. One introduction led to another, and soon Mukherjee found himself at a USC Games showcase festival, meeting a Magnopus recruiter. Within weeks, he was part of the team.
Mukherjee’s fascination with AI began early. At age 10, he encountered the concept in a science fiction encyclopedia that predicted AI would become reality by 2040 or 2050. By early 2023, while completing his undergraduate degree in artificial intelligence at NMIMS University in Mumbai, he watched as professors began experimenting with ChatGPT in their classrooms.
“That moment ignited something enduring: an interest in how AI could serve as an educational partner,” he said.
Suvaditya Mukherjee (right) with Peter Norvig (left), former USC assistant professor and Google director of research.
At ICT, under Nye, Mukherjee developed automated workflows for generating educational content using large language models. The work feeds into AIRCOEE, a collaborative effort with the U.S. Department of Defense to democratize AI education. He’s also presented research twice at the PyTorch Conference in San Francisco — “the world’s premier event dedicated to the framework powering today’s most groundbreaking AI innovations.” The topics ranged from optimizing 3D generation models in 2024 to a poster on the current landscape of 3D generation earlier this year.
His heroes include Geoffrey Hinton, one of the founding fathers of AI and a Turing Prize winner, whom Mukherjee has personally interviewed. Hinton shared stories about his unconventional path to AI — he’s actually a psychologist by training who sought the aid of machine learning while searching for a cure for his wife’s cancer.
“I found that very inspiring,” Mukherjee said.
Looking ahead, Mukherjee envisions working at a frontier AI research lab with access to massive computational resources. His moonshot goals include advancing 3D generation technology to create interactive holographic displays — “a very Iron Man Jarvis-y thing” — and contributing to research on how AI systems actually think.
For now, he’s balancing his coursework at USC Viterbi with his part-time work as a R&D engineer at Magnopus, working in mixed reality and AI. Next semester, he hopes to continue his research with Bolas in the Mixed Intelligence Lab, a new USC lab to explore “the intersections between MR and AI.” All of this, of course, while exploring photography, cooking Indian comfort food, and discovering the cultural landscape of Los Angeles. He’s a Google Developer Expert in Machine Learning who keeps five to seven browser windows open, each with 25 to 30 tabs, as he juggles multiple projects and collaborations.
As for “The Wizard of Oz at Sphere,” Mukherjee takes pride in having his name appear in the credits of a project that pushes the boundaries of both cinema and technology.
For his part, Mukherjee would love to see Christopher Nolan’s “Interstellar” at the Sphere. After traversing the Pacific, voyaging to Oz, perhaps getting lost in a wormhole doesn’t seem quite so farfetched.
Editor’s note: Mukherjee was joined by other USC alumni on “The Wizard of Oz” project at Magnopus, including Gabrielle Roberts, AI creative supervisor, who earned her B.S. in computer science and film and television in 2019; and Melanie Zarzur, product manager, who oversaw the generative AI team and earned her B.S. in civil engineering in 2020.
Published on December 4th, 2025
Last updated on December 4th, 2025
